Coda File System

Re: HTTP-server replication

From: Jan Harkes <>
Date: Thu, 6 Jun 2002 12:13:30 -0400
On Thu, Jun 06, 2002 at 12:36:03PM +0300, Jani Puttonen wrote:
> So the actual question it possible with
> coda distributed file system to build a system,
> where both servers have the same files? So that
> both http-servers have that directory that is
> holding those files as DocumentRoot.

Run a Coda client on both servers, and set the DocumentRoot to
/coda/html or something.

> Has someone tried this kind of system with coda
> before?
> Or is there some better or easier way to do this?

:) our webserver is running like that. Except that I'm not using
'failover' features. It is just a convenient thing to be able to update
the webpages at home or on the road and have the new pages automatically
get propagated to the webserver whenever I get back at CMU.

It also allows for some additional security. Nothing on the webserver
(i.e. not even root) is allowed to write anywhere in /coda, it is
read-only access so there are no tokens stored or used on the machine at
all. As a result, even if someone hacks the webserver, he cannot change
the original html pages. The only thing they can do is change the apache
configuration file and create an alternative document root, which is
easily restored from the read-only backup in Coda.

> This "solution" came to mind. I have
> one coda server and two clients. The clients
> work as http-servers, apache is running on both 
> and the /coda/*** directory is the DocumentRoot.
> I mount the same volumes to both clients.
> The bad thing about this is that I should have
> three computers...and that is waste of resources.

But better for security purposes. However, you can still run the Coda
server on one of the two webservers. If the webserver goes down the
other server will simply continue to run in disconnected mode.

Received on 2002-06-06 12:15:05